EP0579356B1 - Optisches Gerät zur Informationsverarbeitung - Google Patents

Optisches Gerät zur Informationsverarbeitung Download PDF

Info

Publication number
EP0579356B1
EP0579356B1 EP93303623A EP93303623A EP0579356B1 EP 0579356 B1 EP0579356 B1 EP 0579356B1 EP 93303623 A EP93303623 A EP 93303623A EP 93303623 A EP93303623 A EP 93303623A EP 0579356 B1 EP0579356 B1 EP 0579356B1
Authority
EP
European Patent Office
Prior art keywords
spatial light
light modulator
processor
liquid crystal
elements
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Lifetime
Application number
EP93303623A
Other languages
English (en)
French (fr)
Other versions
EP0579356A1 (de
Inventor
Michael Geraint Robinson
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sharp Corp
Original Assignee
Sharp Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sharp Corp filed Critical Sharp Corp
Publication of EP0579356A1 publication Critical patent/EP0579356A1/de
Application granted granted Critical
Publication of EP0579356B1 publication Critical patent/EP0579356B1/de
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/067—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using optical means

Definitions

  • the present invention relates to an optical information processor.
  • a processor may be used, for instance, in neural networks or other applications where matrix-vector multiplications are required.
  • Neural networks are generally based on matrix-vector multiplications in which "interconnection weights" are defined by the elements of a matrix. Such networks have various applications, for instance in the field of pattern recognition, where conventional data processors have proved to be ineffective or inefficient.
  • Neural networks have to be trained for a particular application by applying input data and varying the interconnection weights so as to achieve desired output data.
  • the input data used for training represent a subset of the input data which the neural network will be expected to process correctly when trained.
  • the network comprises three spatial light modulators (SLMs) together with optics and light sources.
  • SLMs spatial light modulators
  • the first SLM is optically addressed and contains the interconnection weights.
  • the second SLM can be addressed optically or electrically and contains the input vector.
  • the third SLM is optically addressed and stores the result of thresholding the vector-matrix product of the input vector and the weight matrix.
  • the result of the computation is optically fed back and multiplied by the input vector.
  • the result of this multiplication is optically applied to the first SLM so as to update the interconnection weights.
  • updating takes place optically in parallel.
  • This arrangement is of limited application because the output stored in the third SLM can only be read optically. Further, this arrangement is only capable of performing unsupervised training in that it cannot be trained to provide a predetermined or desired output in response to a specified input.
  • an optical information processor as defined in the appended Claim 1.
  • optical information processor which is capable of supervised training and in which updating of interconnection weights can be performed effectively in parallel.
  • training time is greatly reduced and is substantially independent of the number of interconnection weights.
  • Use of an optically addressed spatial light modulator for the interconnection weights removes or greatly reduces optical alignment problems during manufacture of the processor. By arranging for read and update paths to correspond optically with each other and by using an optically addressed spatial light modulator in which the optical attenuation cells are defined by light which is incident on the modulator, self-alignment can be achieved.
  • the known type of processor illustrated in Figure 1 comprises an input device 1 in the form of a spatial light modulator for encoding an input vector as a plurality of cells whose light attenuation properties correspond to the components of the input vector.
  • Such components may, for instance, be binary digits or analogue values.
  • Optical radiation such as visible light or infrared radiation, in the form of a collimated beam passes through the modulator 1 and via suitable optics or micro optics 2 onto a two-dimensional spatial light modulator 3 whose optical attenuation cells represent the values of the elements of a weight matrix.
  • Light modulated by each of the elements of the input vector in the modulator 1 is "fanned out" so as to pass through, for instance, a row of cells of the modulator 3.
  • the modulated light from the modulator 3 then passes through further optics 4 and onto an output transducer 5.
  • the light or other optical radiation is "fanned in” to the transducer 5 in a direction substantially perpendicular to the fanning out from the modulator 1 to the modulator 3 so that the transducer 5 produces electrical output signals representing the elements of an output vector which represents the product of the input vector and the weight or interconnection matrix.
  • the interconnection weights of the matrix represented by the spatial light modulator 3 are updated serially during the training phase of the processor when used, for instance, as a neural network.
  • this updating of matrix elements requires considerable time and is substantially equal to the product of the number of matrix elements and the time required for updating each of these elements.
  • accurate optical alignment of the devices shown in Figure 1 is necessary in order to ensure correct operation of the processor. Manufacturing difficulties and cost of manufacture of the processor are therefore substantial.
  • the optical information processor shown in Figure 2 differs from that shown in Figure 1 in several ways.
  • the order of the devices is changed such that, during processing, the optical radiation flows through the spatial light modulator 3 representing the interconnection weight matrix and via the optics 4 to the spatial light modulator 1 representing the input vector. Because of the commutative nature of the vector-matrix multiplication, the order of the devices 1 and 3 can be reversed without affecting the information processing.
  • the modulated optical radiation from the modulator 1 then passes through the optics or micro optics 2 and to the output device 5 which includes a transducer.
  • Updating of the matrix elements represented by the modulator 3 is achieved optically by means of collimated optical radiation flowing in the reverse direction.
  • the device 5 calculates the error, for instance as the difference between the output vector and a target vector.
  • the elements of the resulting error vector are used to control a spatial light modulator forming part of the device 5 so that the error vector modulates the optical radiation during updating.
  • the modulated radiation passes through the modulator 1 so as to form the product of the error vector and the input vector, and this is applied to the modulator 3 where it is used to update the matrix elements in parallel.
  • the modulator 3 is optically addressed and updating of each cell is performed in accordance with the amount of radiation incident on the cell during updating.
  • the elements of the interconnection matrix are updated in parallel so that the time required for each update operation is substantially equal to the update time for a single cell irrespective of the number of cells forming the two-dimensional matrix.
  • no special alignment constraints are placed on the modulator 3 since the individual cells are defined by the light flow during processing and updating and the light flows follow reciprocal paths.
  • complexity and cost of manufacture are greatly reduced compared with a known processor of the type shown in Figure 1.
  • Figure 3 corresponds to that illustrated diagrammatically in Figure 2 but illustrates a modified and possibly more practical arrangement for the optical paths for processing or "read" and update beams. Although collimated beams are required, these are illustrated diagrammatically as single ray paths in Figure 3.
  • the optics 2 and 4 comprise beam splitters for the update beam and the read beam, respectively.
  • the arrangement of the modulators 1 and 3 and the device 5 are otherwise as shown in Figure 2.
  • the read beam 10 is directed into the beam splitter 4, for instance of the split prism type, and is reflected onto the spatial light modulator 3 representing the interconnection weight matrix.
  • the read beam is modulated by the modulator 3 and reflected back along the same ray path and through the beam splitter 4 onto the input vector spatial light modulator 1.
  • Light modulated by the modulator 1 then passes directly through the beam splitter 2 and onto the device 5, which produces the output vector corresponding to the product of the input vector and weight matrix represented by the modulators 1 and 3, respectively.
  • the device 5 subtracts the output vector from a target vector so as to produce an error vector which is applied to an error spatial light modulator disposed on the surface of the device 5 facing the beam splitter 2.
  • the update beam 11 enters the beam splitter 2 and is directed onto the error spatial light modulator where it is modulated by the error vector and reflected back through the beam splitter 2 to the input vector modulator 1.
  • the update beam then passes through the beam splitter 4 and is incident on the spatial light modulator 3.
  • the light received by the modulator 3 thus represents the vector product of the error vector and the input vector and is used to update the weight matrix elements.
  • the spatial light modulator 3 is an optically addressed ferroelectric liquid crystal spatial light modulator combined with a layer of amorphous silicon and suitable electrodes to permit updating to be performed.
  • the regions of the spatial light modulator are thus not committed or predefined by the construction. Instead, the individual cells representing the matrix elements are defined by the light which is incident on the device 3. Thus, no special alignment measures are needed, the device 3 essentially being self-aligning with the other devices of the processor.
  • the update beam modulated to form the error matrix is incident on the amorphous silicon layer and causes electrons to be released in those regions which are exposed to radiation.
  • electrons are injected into the liquid crystal to increment or decrement the light attenuation values of the cells which have been defined and addressed optically.
  • the input spatial light modulator 1 comprises a ferroelectric liquid crystal display comprising a plurality of elongate parallel elements corresponding to the components of the input vector. These elements are aligned parallel to a first dimension of the two dimensional spatial light modulator forming the device 3.
  • the device 5 also comprises a ferroelectric liquid crystal display and an optoelectric transducer having elongate light modulating and transducing elements aligned parallel to the other dimension of the device 3.
  • Each of the devices 3 and 5 may be formed as an integral hybrid device using various known techniques. For instance, the transducer cells and electronics of the device 5 may be formed by means of very large scale integrated circuit technology with the ferroelectric liquid crystal display being subsequently formed on one surface by means of hybrid techniques.
  • Figure 5 illustrates diagrammatically part of a processor of the same type as that shown in Figures 3 and 4 but arranged to perform matrix-matrix multiplication.
  • the devices 1 and 5 are essentially one-dimensional and are used for representing or reading vectors.
  • these devices are replaced by two-dimensional devices so that the input spatial light modulator 21 represents an input matrix as a two-dimensional array of light-attenuating cells.
  • the output device 25 likewise comprises a two-dimensional array of spatial light modulator cells and optoelectric transducer elements.
  • the devices 21 and 25 are separated by a lenslet array 26 comprising a plurality of converging lenses, each of which corresponds to and is aligned with one of the cells of the input modulator 21 and corresponds to and is aligned with a sub-array of the device 25.
  • the lenslet array 26 effectively images the sub-array via the modulator 21 so that each element of the modulator 21 corresponds to and is aligned with a sub-array of cells of the modulator 23.
  • the read and update beams 30 and 31, respectively, are shown as idealised beams in Figure 5 to illustrate that the optical data flows correspond to those illustrated in Figure 2.
  • the devices 21, 25, and 26 require aligning during manufacture, the optically addressed spatial light modulator 23 is effectively self-aligning so that manufacture of the processor is simplified.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Neurology (AREA)
  • Liquid Crystal (AREA)
  • Optical Modulation, Optical Deflection, Nonlinear Optics, Optical Demodulation, Optical Logic Elements (AREA)

Claims (18)

  1. Optischer Informationsprozessor mit einem ersten Raumlichtmodulator (3, 23), der Verbindungsgewichtungsdaten aufnehmen soll, einem zweiten Raumlichtmodulator (1, 21), der Eingangsdaten aufnehmen soll, und einem dritten Raumlichtmodulator (5, 25), wobei der erste Raumlichtmodulator (3, 23) optisch adressierbar ist und der erste, zweite und dritte Raumlichtmodulator (1, 3, 5, 21, 23, 25) einen optischen Aktualisierungspfad zum Aktualisieren von Gewichtungsdaten aus dem dritten Raumlichtmodulator (5, 25) über den zweiten Raumlichtmodulator (1, 21) zum ersten Raumlichtmodulator (3, 23) bilden, dadurch gekennzeichnet, dass der dritte Raumlichtmodulator (5, 25) ferner eine optoelektrische Wandlereinrichtung, die mit dem ersten und zweiten Raumlichtmodulator (1, 3, 21, 23) zusammenwirkt, um einen optischen Verarbeitungspfad vom ersten Raumlichtmodulator (3, 23) zum zweiten Raumlichtmodulator (1, 21) zur optoelektrischen Wandlereinrichtung zu bilden, eine Verarbeitungseinrichtung zum Verarbeiten von Ausgangsdaten von der optoelektrischen Wandlereinrichtung sowie vorbestimmter Sollausgangsdaten zum Erzeugen von Abweichungsdaten, und eine Steuereinrichtung zum Steuern des dritten Raumlichtmodulators so, dass er während der Gewichtungsaktualisierung Abweichungsdaten enthält, aufweist.
  2. Prozessor nach Anspruch 1, dadurch gekennzeichnet, dass die Verarbeitungseinrichtung so ausgebildet ist, dass sie die Abweichungsdaten als Differenz zwischen den Ausgangsdaten und den vorbestimmten Sollausgangsdaten berechnet.
  3. Prozessor nach einem der Ansprüche 1 oder 2, dadurch gekennzeichnet, dass der zweite Raumlichtmodulator (1, 21) eine erste Flüssigkristallvorrichtung ist.
  4. Prozessor nach Anspruch 3, dadurch gekennzeichnet, dass die erste Flüssigkristallvorrichtung (1, 21) eine ferroelektrische Flüssigkristallvorrichtung ist.
  5. Prozessor nach einem der vorstehenden Ansprüche, dadurch gekennzeichnet, dass der dritte Raumlichtmodulator (5, 25) eine zweite Flüssigkristallvorrichtung ist.
  6. Prozessor nach Anspruch 5, dadurch gekennzeichnet, dass die zweite Flüssigkristallvorrichtung eine ferroelektrische Flüssigkristallvorrichtung ist.
  7. Prozessor nach einem der vorstehenden Ansprüche, dadurch gekennzeichnet, dass der erste Raumlichtmodulator (3, 23) eine dritte Flüssigkristallvorrichtung ist.
  8. Prozessor nach Anspruch 7, dadurch gekennzeichnet, dass die dritte Flüssigkristallvorrichtung eine ferroelektrische Flüssigkristallvorrichtung ist.
  9. Prozessor nach Anspruch 8, dadurch gekennzeichnet, dass die dritte Flüssigkristallvorrichtung eine Photoelektronenquelle und eine Vorspannungseinrichtung zum selektiven Injizieren von Elektronen von der Quelle in die ferroelektrische Flüssigkristallvorrichtung aufweist.
  10. Prozessor nach Anspruch 9, dadurch gekennzeichnet, dass die Photoelektronenquelle amorphes Silicium aufweist.
  11. Prozessor nach einem der Ansprüche 9 oder 10, dadurch gekennzeichnet, dass die Vorspannungseinrichtung eine Einrichtung zum Anlegen einer Potentialdifferenz an den ferroelektrischen Flüssigkristall der Vorrichtung aufweist.
  12. Prozessor nach einem der vorstehenden Ansprüche, dadurch gekennzeichnet, dass der erste Raumlichtmodulator (3, 23) ein zweidimensionales Array von Elementen aufweist.
  13. Prozessor nach Anspruch 22, dadurch gekennzeichnet, dass der zweite Raumlichtmodulator (1) eine Vielzahl langgestreckter, paralleler Elemente aufweist.
  14. Prozessor nach einem der Ansprüche 12 oder 13, dadurch gekennzeichnet, dass der dritte Raumlichtmodulator (5) eine Vielzahl langgestreckter, paralleler Elemente aufweist.
  15. Prozessor nach einem der Ansprüche 12 bis 14, dadurch gekennzeichnet, dass die Wandlereinrichtung eine Vielzahl langgestreckter, paralleler Wandlerelemente aufweist.
  16. Prozessor nach Anspruch 12, dadurch gekennzeichnet, dass der zweite Raumlichtmodulator (21) ein zweidimensionales Array von Elementen aufweist, von denen jedes einer Vielzahl der Elemente des ersten Raumlichtmodulators (23) aufweist.
  17. Prozessor nach Anspruch 16, dadurch gekennzeichnet, dass der dritte Raumlichtmodulator (25) ein zweidimensionales Array von Elementen aufweist, wobei jedes Element des zweiten Raumlichtmodulators (21) einer Vielzahl der Elemente des dritten raumlichtmodulators entspricht.
  18. Prozessor nach einem der Ansprüche 16 oder 17, dadurch gekennzeichnet, dass die Wandlereinrichtung ein zweidimensionales Array von Wandlerelementen aufweist, wobei jedes Element des zweiten Raumlichtmodulators (21) einer Vielzahl der Wandlerelemente entspricht.
EP93303623A 1992-05-11 1993-05-11 Optisches Gerät zur Informationsverarbeitung Expired - Lifetime EP0579356B1 (de)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
GB9210080 1992-05-11
GB9210080A GB2267165A (en) 1992-05-11 1992-05-11 Optical information processor

Publications (2)

Publication Number Publication Date
EP0579356A1 EP0579356A1 (de) 1994-01-19
EP0579356B1 true EP0579356B1 (de) 1997-09-24

Family

ID=10715298

Family Applications (1)

Application Number Title Priority Date Filing Date
EP93303623A Expired - Lifetime EP0579356B1 (de) 1992-05-11 1993-05-11 Optisches Gerät zur Informationsverarbeitung

Country Status (5)

Country Link
US (1) US5383042A (de)
EP (1) EP0579356B1 (de)
JP (1) JP3707804B2 (de)
DE (1) DE69314107T2 (de)
GB (1) GB2267165A (de)

Families Citing this family (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AT400371B (de) * 1994-03-02 1995-12-27 Budil Matthias Dr Optischer vektormultiplizierer für neuronale netze
ATE187825T1 (de) * 1994-03-02 2000-01-15 Matthias Budil Optischer vektormultiplizierer für neuronale netze
JPH09230954A (ja) * 1996-02-28 1997-09-05 Olympus Optical Co Ltd ベクトル規格化装置
GB2341965A (en) * 1998-09-24 2000-03-29 Secr Defence Pattern recognition
US6115168A (en) * 1998-10-29 2000-09-05 Advanced Optical Technologies, Inc. Integrated optical retroreflecting modulator
AU6342700A (en) * 1999-07-09 2001-01-30 Opts, Inc. Adaptive compressive network
US6643843B1 (en) * 1999-11-23 2003-11-04 Ellenby Technologies Inc. Methods and apparatus for optical communication update of program memory in embedded systems
JP2003332560A (ja) * 2002-05-13 2003-11-21 Semiconductor Energy Lab Co Ltd 半導体装置及びマイクロプロセッサ
JP4094386B2 (ja) * 2002-09-02 2008-06-04 株式会社半導体エネルギー研究所 電子回路装置
JP4373063B2 (ja) 2002-09-02 2009-11-25 株式会社半導体エネルギー研究所 電子回路装置
JP4574118B2 (ja) * 2003-02-12 2010-11-04 株式会社半導体エネルギー研究所 半導体装置及びその作製方法
EP1767584A4 (de) * 2004-07-14 2010-03-31 Kaneka Corp Herstellungsverfahren für eine härtbare harzzusammensetzung
US8775341B1 (en) 2010-10-26 2014-07-08 Michael Lamport Commons Intelligent control with hierarchical stacked neural networks
US9015093B1 (en) 2010-10-26 2015-04-21 Michael Lamport Commons Intelligent control with hierarchical stacked neural networks
US9459451B2 (en) 2013-12-26 2016-10-04 Microsoft Technology Licensing, Llc Eye tracking apparatus, method and system
US9377623B2 (en) * 2014-08-11 2016-06-28 Microsoft Technology Licensing, Llc Waveguide eye tracking employing volume Bragg grating
US9494799B2 (en) 2014-09-24 2016-11-15 Microsoft Technology Licensing, Llc Waveguide eye tracking employing switchable diffraction gratings
US10217023B1 (en) * 2017-06-14 2019-02-26 The United States Of America As Represented By Secretary Of The Navy Image recognition system using a programmable photonic neural network
CN114912604A (zh) * 2018-05-15 2022-08-16 轻物质公司 光子计算系统和光学执行矩阵向量乘法的方法
TW202032187A (zh) 2018-06-04 2020-09-01 美商萊特美特股份有限公司 實數光子編碼
TW202111467A (zh) 2019-02-25 2021-03-16 美商萊特美特股份有限公司 路徑數平衡式通用光子網路
US10803259B2 (en) 2019-02-26 2020-10-13 Lightmatter, Inc. Hybrid analog-digital matrix processors
KR20220039775A (ko) 2019-07-29 2022-03-29 라이트매터, 인크. 선형 포토닉 프로세서를 이용한 아날로그 계산을 위한 시스템들 및 방법들
US11093215B2 (en) 2019-11-22 2021-08-17 Lightmatter, Inc. Linear photonic processors and related methods
EP4186005A1 (de) 2020-07-24 2023-05-31 Lightmatter, Inc. Systeme und verfahren zur verwendung photonischer freiheitsgraden in einem photonischen prozessor

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5004309A (en) * 1988-08-18 1991-04-02 Teledyne Brown Engineering Neural processor with holographic optical paths and nonlinear operating means
JPH02143391A (ja) * 1988-11-25 1990-06-01 Ricoh Co Ltd 並列光情報処理装置
US4937776A (en) * 1988-12-13 1990-06-26 Nippon Sheet Glass Co., Ltd. Matrix-vector multiplication apparatus
US4959532A (en) * 1989-02-16 1990-09-25 Hughes Aircraft Company Optical neural network and method
US4963725A (en) * 1989-10-24 1990-10-16 Rockwell International Corporation Adaptive optical neural network
US5235440A (en) * 1989-11-06 1993-08-10 Teledyne Industries, Inc. Optical interconnector and highly interconnected, learning neural network incorporating optical interconnector therein
WO1991007714A1 (en) * 1989-11-20 1991-05-30 British Telecommunications Public Limited Company Neural networks
GB2245732A (en) * 1990-06-29 1992-01-08 Philips Electronic Associated Optical data processing device

Also Published As

Publication number Publication date
EP0579356A1 (de) 1994-01-19
JPH0643957A (ja) 1994-02-18
DE69314107T2 (de) 1998-03-26
GB9210080D0 (en) 1992-06-24
GB2267165A (en) 1993-11-24
JP3707804B2 (ja) 2005-10-19
DE69314107D1 (de) 1997-10-30
US5383042A (en) 1995-01-17

Similar Documents

Publication Publication Date Title
EP0579356B1 (de) Optisches Gerät zur Informationsverarbeitung
US5004309A (en) Neural processor with holographic optical paths and nonlinear operating means
US4187000A (en) Addressable optical computer and filter
US5321639A (en) Dual-scale topology optoelectronic matrix algebraic processing system
US5784309A (en) Optical vector multiplier for neural networks
US5220643A (en) Monolithic neural network element
US5132813A (en) Neural processor with holographic optical paths and nonlinear operating means
EP0260281B1 (de) Optische datenverarbeitungsanordnungen und verfahren zur matrizen-inversion, -multiplikation und -addition
US4595994A (en) Optical engagement array multiplication
Arsenault et al. An introduction to optics in computers
US4888724A (en) Optical analog data processing systems for handling bipolar and complex data
US5113485A (en) Optical neural network system
Athale et al. Compact architectures for adaptive neural nets
US5487026A (en) Multiplying device, linear algebraic processor, neuromorphic processor, and optical processor
Kane A smart pixel-based feedforward neural network
Owechko et al. Holographic neurocomputer utilizing laser diode light source
Ansari et al. Optical computing
Saxena et al. Adaptive Multilayer Optical Neural Network Design
Robinson et al. Custom electro-optic devices for optically implemented neuromorphic computing systems
Liu Optical implementation of inner-product neural associative memory
Ruiz-Llata et al. Design of a compact neural network based on an optical broadcast architecture
JPH0259916A (ja) 光演算装置
Berry et al. Guided Wave Vector-Matrix Multiplier
Chao et al. Optoelectronic implementation of neocognitron for pattern recognition
Cheng Variable-frequency cross-grating interferometer

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): DE FR GB

17P Request for examination filed

Effective date: 19940330

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

17Q First examination report despatched

Effective date: 19961213

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

GRAA (expected) grant

Free format text: ORIGINAL CODE: 0009210

AK Designated contracting states

Kind code of ref document: B1

Designated state(s): DE FR GB

REF Corresponds to:

Ref document number: 69314107

Country of ref document: DE

Date of ref document: 19971030

ET Fr: translation filed
PLBE No opposition filed within time limit

Free format text: ORIGINAL CODE: 0009261

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT

26N No opposition filed
REG Reference to a national code

Ref country code: GB

Ref legal event code: IF02

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: DE

Payment date: 20060508

Year of fee payment: 14

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: GB

Payment date: 20060510

Year of fee payment: 14

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: FR

Payment date: 20060515

Year of fee payment: 14

GBPC Gb: european patent ceased through non-payment of renewal fee

Effective date: 20070511

REG Reference to a national code

Ref country code: FR

Ref legal event code: ST

Effective date: 20080131

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: DE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20071201

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: GB

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20070511

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: FR

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20070531